Guillaume Pernelle

Lausanne, Switzerland · audiala.com

• Building the AI travel companion: personal audio tours for any city — planned on the web, narrated on the street, fully offline.
• 1,000+ cities live in 11 languages; 50,000 proprietary narration scripts; a new city costs ~$10 to produce where legacy providers charge destinations $10,000+.
• Built the entire stack solo: iOS, Android, web app, AI agent harness, and a 680,000-page SEO engine cited by the LLMs themselves.
• Microsoft AI Accelerator; cloud grants from Microsoft, AWS, and Google.

Switzerland · hyll.com

• Led the implementation of large language models in the company's infrastructure, enabling new products and services.
• Created an AI chatbot and a new recommendation system.
• Trained the team on the latest advances in machine learning and natural language processing.

Remote, Switzerland · joined through the acquisition of Magic Carpet AI · blockchain.com

• Led the integration of Magic Carpet AI's technology into Blockchain.com's trading stack following the acquisition of our company.
• Modelled order-book dynamics to cut execution slippage and market impact — scaled trading volumes 4×.
• Improved Blockchain.com's infrastructure — deployment ease and procedures, trading latencies reduced by up to 50×.
• Built data pipelines on GCP (BigQuery, BigTable, Airflow); researched order-book and tick-data signals for execution algorithms and new strategies.

London, UK · acquired by Blockchain.com, Nov 2021 (announcement, advisor note)

• Co-founded Magic Carpet AI: deep reinforcement learning applied to portfolio management — new strategy classes optimizing risk/return across asset baskets, taken from research to live capital.
• As CTO, led research and development for four years with an 8-person London team of quant traders and data scientists.
• Developed a family of stable, high-performing investment algorithms that learn from past conditions and adapt to current ones to balance risks and returns.
• Led the development of a complete in-house trading and risk-management system tailored for digital assets.
• Eight-figure acquisition by Blockchain.com, November 2021.

Geneva, Switzerland · testing and quality-control solutions · qmt.ch

• Built automated vision-control solutions: data acquisition, signal and image processing, algorithm and GUI development, on-site installation.
• Analyzed customer requirements and co-designed technical solutions across optics, software, and hardware; owned testing, validation, and technical documentation.
• Led QMTSubface, a partnership with EPFL applying machine learning to qualify the aesthetic properties of objects.

Surgical Planning Laboratory, Brigham and Women's Hospital, Boston, USA · spl.harvard.edu

Master's thesis on medical-image segmentation and registration using mechanically inspired models — evaluation phantom design, and efficient, publicly released software in Python and C++. Followed by several peer-reviewed publications (MICCAI 2013, Medical Image Analysis 2017).

Munich, Germany · optcon.de

Built OptCon's website and ran its search-engine optimization.

College Station, Texas, USA · corinnova.com

Prototyped analog hardware reading an electrocardiogram and driving a pump, for the world's first minimally invasive non-blood-contacting cardiac assist device — the heart wall is pushed in sync with the ECG.

Toulouse, France

Assembly of motors, winglets, and landing gear; verification of electrical systems.

Bioengineering Department, Computational Neuroscience Laboratory · supervisor Dr. Claudia Clopath

Modeled the dynamics of spiking recurrent neural networks — how gap-junction plasticity regulates network-wide oscillations and ensures robust information transfer between brain regions; also studied the relationship between blood oxygen concentration and neural activity. Graduate Teaching Assistant for Computational Neuroscience, Mathematics, and Python. PhD thesis.

Mechanical Engineering · major Medical Technology, minors Micro Technology & Product Development · T.I.M.E. double-degree program

Thesis at the Surgical Planning Laboratory, Harvard Medical School — research, software engineering, and mechanical engineering applied to radiology.

Major Image & Signal Processing, minors Management & Communication

Thesis: an imaging system based on light polarization for biomedical applications — detection of skin cancers.

Gap junction plasticity as a mechanism to regulate network-wide oscillations

Validation of catheter segmentation for MR-guided gynecologic cancer brachytherapy

github.com/needlefinder/needlefinder-3d-unet

3D U-Net convolutional model segmenting needles from 3D MRI, built on TensorFlow.

github.com/needlefinder/needlefinder

Open-source extension with interactive tools to segment needles in MR/CT images, validated on MRI from gynecological brachytherapy cases (MICCAI 2013).

github.com/gpernelle/igyne

Procedure and applicator selection, data import, initial and refined applicator registration, needle position planning and detection.